(669b) Ultraviolet Spectroscopy-Based Enhanced Estimation of Monoclonal Antibodies Using Smart Wavelength Selection

Monoclonal antibodies (mAbs) are therapeutic proteins used for the targeted treatment of various diseases, such as autoimmune disorders and cancer. Process economics have motivated the push towards continuous biomanufacturing, in turn motivating increased attention to monitoring and control algorithms. The present study focuses on mAb production using Chinese hamster ovary cells, and specifically on the monitoring of the chromatography-based separation process to separate the mAbs from the harvest cell culture fluid (HCCF). Current mAb monitoring/estimation techniques have used techniques such as Partial Least Squares and Artificial Neural Networks to relate ultraviolet (UV) absorbance to mAb concentrations and implemented these tools in the context of chromatographic separation^1,2. The concentration of HCCF species shifts throughout the course of a perfusion bioreactor run impacting the background absorbance measured by UV spectroscopy. The most recent developments of inferential sensors for monitoring mAb concentrations using UV absorbance include real-time permeate specific baseline subtraction, using conductivity measurements to determine the baseline³. Existing results have focused on offline implementation of these techniques, with diode array detectors (DADs) which simultaneously measures the whole UV spectrum. The present work proposes a PLS model based online monitoring tool with dynamic wavelength selection making use of UV absorbance collected through the BioPAT® SpectroUV (a DAD connected to a flow cell). The key idea of the dynamic wavelength selection is inspired from recent results that show that improved estimation is possible through careful selection of spectrometer wavelengths used for estimation⁴. Spectrometers that are limited by the number of wavelengths to measure absorbance (typically three to four) are commonly used in industry. The present work applies this idea dynamically (as measurements are collected online) to change the selected wavelengths for improved estimation. The method is demonstrated on Sartobind® Rapid A Nano membrane adsorbers separating mAb from HCCF for two different samples from a 12 day perfusion run. Using the BioPAT® SpectroUV, UV absorbance data across 201 wavelengths from 190 to 390 nm at 1 nm increments is acquired for multiple breakthrough curves. The mAb concentration of the breakthrough curve samples is measured using the Label-Free Octet® Red96 and Octet® ProA Biosensors. Two benchmarks based on existing methodologies in Literature (univariate Beer-Lambert Law and multivariate modelling using PLS) are evaluated and compared against the proposed dynamic wavelength selection algorithm. Experimental results are provided to demonstrate the improved estimation possible with the proposed techniques.

References:

(1) Rüdt, M.; Brestrich, N.; Rolinger, L.; Hubbuch, J. Real‐time Monitoring and Control of the Load Phase of a Protein A Capture Step. Biotechnol. Bioeng. 2017, 114 (2), 368–373. https://doi.org/10.1002/bit.26078.

(2) Rolinger, L.; Rüdt, M.; Hubbuch, J. Comparison of UV‐ and Raman‐based Monitoring of the Protein A Load Phase and Evaluation of Data Fusion by PLS Models and CNNs. Biotechnol. Bioeng. 2021, 118 (11), 4255–4268. https://doi.org/10.1002/bit.27894.

(3) Rolinger, L.; Rüdt, M.; Hubbuch, J. A Multisensor Approach for Improved Protein A Load Phase Monitoring by Conductivity‐based Background Subtraction of UV Spectra. Biotechnol. Bioeng. 2021, 118 (2), 905–917. https://doi.org/10.1002/bit.27616.

(4) Gough, I. A.; Rassenberg, S.; Velikonja, C.; Corbett, B.; Latulippe, D. R.; Mhaskar, P. Selective Protein Quantification on Continuous Chromatography Equipment with Limited Absorbance Sensing: A Partial Least Squares and Statistical Wavelength Selection Solution. J. Chemom. 2024, 38 (7), e3541. https://doi.org/10.1002/cem.3541.